电子学报 ›› 2020, Vol. 48 ›› Issue (5): 975-984.DOI: 10.3969/j.issn.0372-2112.2020.05.019

• 学术论文 • 上一篇    下一篇

融合混沌优化和改进模糊聚类的图像分割算法

朱占龙1,2,3, 刘永军1,3   

  1. 1. 河北地质大学信息工程学院, 河北石家庄 050031;
    2. 河北省光电信息与地球探测技术重点实验室, 河北石家庄 050031;
    3. 河北地质大学人工智能与机器学习研究室, 河北石家庄 050031
  • 收稿日期:2019-03-11 修回日期:2019-10-27 出版日期:2020-05-25 发布日期:2020-05-25
  • 作者简介:朱占龙 男,1984年5月出生,河北石家庄人.分别于2008年和2011年在燕山大学获得学士学位、硕士学位,2015年于东南大学获得博士学位.现为河北地质大学信息工程学院讲师,主要研究方向为图像处理.E-mail:zhuzl@hgu.edu.cn;刘永军 男,1970年9月出生,河北石家庄人.副教授、中国计算机学会高级会员.1992年于哈尔滨工业大学获得学士学位,2013年于石家庄铁道大学获得硕士学位.现为河北地质大学信息工程学院教师,主要从事计算机专业教育、智能处理研究.
  • 基金资助:
    河北省高等学校科学技术研究(No.BJ2018029);河北省教育厅重点(No.ZD2018212);河北地质大学博士科研启动基金(No.BQ201606)

A Novel Algorithm by Incorporating Chaos Optimization and Improved Fuzzy C-Means for Image Segmentation

ZHU Zhan-long1,2,3, LIU Yong-jun1,3   

  1. 1. School of Information Engineering, Heibei GEO University, Shijiazhuang, Hebei 050031, China;
    2. Hebei Key Laboratory of Optoelectronic Information and Geo-Detection Technology, Shijiazhuang, Hebei 050031, China;
    3. Laboratory of Artificial Intelligence and Machine Learning, Heibei GEO University, Shijiazhuang, Hebei 050031, China
  • Received:2019-03-11 Revised:2019-10-27 Online:2020-05-25 Published:2020-05-25

摘要: 基于邻域广义模糊聚类算法能够分割含噪声灰度图像,但是如果图像灰度分布不均衡或者起始的聚类中心设置不合适仍会导致该算法分割失败,为此,提出一种基于混沌优化和改进模糊聚类算法相融合的图像分割算法.首先,将每一类的隶属度之和引入基于邻域广义模糊聚类算法的目标函数中,从而能够均衡较大类和较小类对目标函数的贡献.其次,以新目标函数为基础,利用拉格朗日乘子法推导出相应的隶属度和聚类中心.再次,将混沌优化和改进模糊聚类算法联合得到最优解,即最合适的聚类中心,细节上,每一代的聚类中心分别由混沌系统和改进模糊聚类算法两种路径产生,具有较小目标函数的聚类中心进入下一个迭代进程.最后,利用具有不平衡特性的无损检测图像进行实验,结果表明本文算法具有更高的分割准确率和更好的视觉效果.

关键词: 图像分割, 混沌优化, 模糊聚类, 灰度分布不均衡

Abstract: The spatial generalized fuzzy c-means clustering algorithm (GFCM_S) is a popular technique for image segmentation,but it is not so effective when the image has the features of unequal cluster sizes or the initial cluster centers we choose are improper.In this paper,for solving the above shortcomings of GFCM_S,a novel algorithm incorporating chaos optimization and improved fuzzy c-means (CIGFCM_S) is proposed.Firstly,each size of clusters is integrated into the objective function of GFCM_S so as to equalize the contribution of larger and smaller clusters to the objective function.Secondly,the iteratively membership degree and cluster centers are deduced by the Lagrange multiplier method.Thirdly,a new iterative strategy is used to seek the optimal solutions.In detail,the optimal solutions of next generation are searched by two-paths,one path originates chaos optimization and the other is obtained by updating membership degree and cluster centers on the basis of current optimal solutions,and then the better solutions go to next generation until the end.Lastly,the non-destructive testing (NDT) images with the characters of unequal cluster sizes are used for experiments,the results show that the proposed algorithm has better segmentation accuracy and visual effects.

Key words: image segmentation, chaos optimization, fuzzy clustering, unequal cluster sizes

中图分类号: